Foot Ulcers: Characterising the Microbiome of Ulcers Karen Smith1, Andrew Collier2, Eleanor M

Smith et al. BMC Microbiology (2016) 16:54 DOI 10.1186/s12866-016-0665-z

RESEARCH ARTICLE Open Access One step closer to understanding the role of bacteria in diabetic foot ulcers: characterising the microbiome of ulcers Karen Smith1, Andrew Collier2, Eleanor M. Townsend1,3, Lindsay E. O’Donnell3, Abhijit M. Bal2, John Butcher1, William G. Mackay1, Gordon Ramage3* and Craig Williams1

Abstract Background: The aim of this study was to characterise the microbiome of new and recurrent diabetic foot ulcers using 16S amplicon sequencing (16S AS), allowing the identification of a wider range of bacterial species that may be important in the development of chronicity in these debilitating wounds. Twenty patients not receiving antibiotics for the past three months were selected, with swabs taken from each individual for culture and 16S AS. DNA was isolated using a combination of bead beating and kit extraction. Samples were sequenced on the Illumina Hiseq 2500 platform. Results: Conventional laboratory culture showed positive growth from only 55 % of the patients, whereas 16S AS was positive for 75 % of the patients (41 unique genera, representing 82 different operational taxonomic units (OTU’s). S. aureus was isolated in 72 % of culture-positive samples, whereas the most commonly detected bacteria in all ulcers were Peptoniphilus spp., Anaerococcus spp. and Corynebacterium spp., with the addition of Staphylococcus spp. in new ulcers. The majority of OTU’s residing in both new and recurrent ulcers (over 67 %) were identified as facultative or strict anaerobic Gram-positive organisms. Principal component analysis (PCA) showed no difference in clustering between the two groups (new and recurrent ulcers). Conclusions: The abundance of anaerobic bacteria has important implications for treatment as it suggests that the microbiome of each ulcer “starts afresh” and that, although diverse, are not distinctly different from one another with respect to new or recurrent ulcers. Therefore, when considering antibiotic therapy the duration of current ulceration may be a more important consideration than a history of healed ulcer. Keywords: Diabetes,Diabeticfootulcer,Infection,Microbiome, Next generation sequencing, 16S amplicon sequencing

Background high plantar pressures [2]. Diabetic foot ulcers have a A serious complication of diabetes is the development of significant negative impact on health and quality of life foot ulcers. Patients with diabetes are believed to have a and are the most common cause of hospitalisation in pa- 12–25 % lifetime risk of developing a foot ulcer [1]. The tients with diabetes [3]. aetiology of diabetic foot ulceration is complex. Foot ul- Diabetic foot ulcers often heal very slowly, due to cers often develop due to a combination of intrinsic fac- diabetes-associated micro-vascular disease and impaired tors, such as peripheral neuropathy, poor extremity host immune response, and these open wounds provide perfusion, foot deformity, changes to the plantar foot a niche for infection [4, 5]. Bacteria can exist within the soft tissues plus extrinsic mechanical factors, such as wound as multi-layered microbial communities, known as biofilms, surrounded by a self-produced protective * Correspondence: [email protected] extracellular ‘slime’ [6]. The presence of a biofilm makes 3 Infection and Immunity Research Group, Glasgow Dental School, School of infections very difficult to resolve, as the structure Medicine, College of Medical, Veterinary and Life Sciences, University of Glasgow, Glasgow, UK shields the encased cells from antimicrobial agents and Full list of author information is available at the end of the article

© 2016 Smith et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Smith et al. BMC Microbiology (2016) 16:54 Page 2 of 12

the host immune system, allowing bacteria to persist Table 1 Clinical laboratory culture of diabetic foot wound and impair healing [7]. Many foot ulcers fail to heal and samples cause serious complications, such as osteomyelitis. Infec- Sample number New or recurrent ulcer Bacteria isolated tion is the most common cause of lower limb amputa- 1 Recurrent NSGa tion in diabetic foot ulcers [3]. In the UK, more than 2 Recurrent NSG one hundred amputations are carried out each week in 3 Recurrent Candida spp. individuals with diabetes [1]. In addition to the significant trauma, the cost to the NHS of treating infected 4 New Mixed growth + anaerobe diabetic foot ulcers and resulting amputations is esti- 5 New Mixed growth + anaerobe mated to be in the region of £900 million per year [8]. 6 Recurrent S. aureus + anaerobe Standard treatment of diabetic foot ulcers involves de- 7 New S. aureus + anaerobe bridement of the necrotic tissue, management of infection 8 New NSG and off-loading of the ulcer [9]. Infection is routinely con- 9 New S. aureus + beta-haemolytic firmed by laboratory culture of bacteria in a swab taken Streptococcus (Group G) from the wound. Culture-dependent methods show bias 10 Recurrent NSG towards microorganisms that are able to grow well on la- 11 New NSG boratory culture media. More fastidious organisms may not be identified, resulting in a delay in the appropriate 12 New S. aureus treatment [10, 11]. Molecular methods are advancing and 13 New S. aureus + beta-haemolytic becoming more accessible and affordable (including 16S Streptococcus (Group G) amplicon sequencing [16S AS]), and it is now possible to 14 Recurrent S. aureus use bacterial DNA from the wound site to identify the 15 Recurrent NSG pathogens present [12]. A greater understanding of both 16 Recurrent NSG the bacteria present in diabetic foot ulcers and how these 17 New S. aureus bacteria interact with one another, and the host, will be 18 New NSG crucial for the development of reliable models of infection and effective treatments. 19 Recurrent S. aureus It was our hypothesis that the specific microbiome as- 20 Recurrent NSG sociated with new and recurrent diabetic foot ulcers dif- aNSG - No significant growth fered, and that this impacted the ability to effectively manage these wounds. Therefore, the aim of this study aureus was confirmed in 50 % of the samples tested by was to use 16S AS technologies to characterise the PCR. Spearman’s rank correlation showed that this cor- microbiome of new and recurrent ulcers and to under- related to culture results (p < 0.05, CI 0.5761–0.9642). take comparative analyses to gain understanding of whether the presence of certain microorganisms were Microbiome analysis diabetic foot ulcers by 16S amplicon associated with the inability of an ulcer to heal. sequencing DNA was successfully amplified and sequenced from 16 of the Results samples (75 %), including 9 from new ulcers and 7 from recur- The detection of bacteria in diabetic foot ulcers by rent ulcers. These 16 samples produced an average of 1,767,142 conventional laboratory culture and molecular methods sequence reads, which were filtered for quality and assigned an By conventional laboratory culture, 11 samples (55 %) OTU using a minimum sequence similarity of 97 %. The abun- were positive for microbial growth by standard aerobic dance of each species in wound samples of greater than 0.5 % and anaerobic culture methods (4 recurrent ulcer sam- was reported. Raw data sets are available from the following ples and 7 new ulcer samples) (Table 1). The remaining website: http://www.ncbi.nlm.nih.gov/bioproject/304940. 9 swab samples (45 %) produced no significant growth In the 16 ulcer samples there were 41 unique genera, (6 recurrent and 3 new ulcer samples). In the samples representing 82 different OTU’s, which ranged from 21 that were positive for bacterial growth, 6 contained more different species (sample 4) to only 2 species (sample 9). than one species. S. aureus was isolated in 8 of culture- The species identified in new and non-healing ulcers are positive samples, anaerobes were isolated from 4 sam- displayed in Table 2, along with their frequency of detec- ples, beta-haemolytic streptococci from 2 samples, and tion. In new ulcers there were 94 different OTU’s Candida spp. was identified alone in only 1 sample. One identified; most frequently detected genera were Pep- control swab sample collected from the healthy skin toniphilus (6 samples), Staphylococcus (5 samples), (sample 19) showed a moderate growth of S. aureus. Anaerococcus (5 samples) and Corynebacterium (4 samples). These results are summarised in Table 1. Moreover, S. In recurrent ulcers 73 unique OTU’s were identified; Smith et al. BMC Microbiology (2016) 16:54 Page 3 of 12

Table 2 List of species identified in the new and recurrent ulcers by 16S AS Species identified No. of samples Gram type Oxygen tolerance New Recurrent Actinobaculum massiliense 1 0 Gram Positive Facultative anaerobe Actinobaculum schaalii 2 2 Gram Positive Facultative anaerobe Actinomyces europaeus 0 1 Gram Positive Facultative anaerobe Actinomyces hominis 0 1 Gram Positive Facultative anaerobe Actinomyces neuii 0 1 Gram Positive Facultative anaerobe Actinomyces radingae 0 1 Gram Positive Facultative anaerobe Alcaligenes faecalis 0 1 Gram Negative Aerobe Anaerococcus murdoch 3 2 Gram Positive Anaerobe Anaerococcus tetradius 1 0 Gram Positive Anaerobe Anaerococcus vaginalis 5 4 Gram Positive Anaerobe Bacteriodes fragilis 2 0 Gram Negative Anaerobe Bilophila wadsworthia 1 0 Gram Negative Anaerobe Bulleidia extructa 1 0 Gram Positive Anaerobe Campylobacter ureolyticus 2 1 Gram Negative Anaerobe Clostridium saccharogumia 2 0 Gram Positive Anaerobe Corynebacterium accolens 0 1 Gram Positive Facultative anaerobe Corynebacterium amycolatum 4 0 Gram Positive Facultative anaerobe Corynebacterium aurimucosum 2 1 Gram Positive Facultative anaerobe Corynebacterium freiburgense 0 1 Gram Positive Facultative anaerobe Corynebacterium hansenii 1 0 Gram Positive Facultative anaerobe Corynebacterium mycetoide 0 1 Gram Positive Facultative anaerobe Corynebacterium simulans 1 3 Gram Positive Facultative anaerobe Corynebacterium tuberculostearicum 1 0 Gram Positive Facultative anaerobe Corynebacterium xerosis 1 0 Gram Positive Facultative anaerobe Dermabacter hominis 2 2 Gram Positive Facultative anaerobe Dialister propionicifaciens 1 0 Gram Negative Anaerobe Dialister micraerophilus 1 0 Gram Negative Anaerobe Dialister pneumosintes 1 0 Gram Negative Anaerobe Eggerthella lenta 1 0 Gram Positive Anaerobe Enterobacter hormaechei 0 2 Gram Negative Facultative anaerobe Enterococcus canintestini 0 2 Gram Negative Facultative anaerobe Escherichia fergusonii 0 1 Gram Negative Facultative anaerobe Escherichia vulneris 0 1 Gram Negative Facultative anaerobe Finegoldia magna 5 5 Gram Positive Anaerobe Fusobacterium canifelinum 1 0 Gram Negative Anaerobe Fusobacterium nucleatum 1 0 Gram Negative Anaerobe Fusobacterium periodontium 1 0 Gram Negative Anaerobe Gemella morbillorum 0 1 Gram Positive Anaerobe Granulicatella adiacens 0 1 Gram Positive Facultative anaerobe Haemophilus parainfluenzae 1 1 Gram Negative Facultative anaerobe Helcococcus kunzii 1 2 Gram Positive Anaerobe Kocuria atrinae 1 Gram Positive Anaerobe Leclercia adecarboxylata 0 2 Gram Negative Facultative anaerobe Smith et al. BMC Microbiology (2016) 16:54 Page 4 of 12

Table 2 List of species identified in the new and recurrent ulcers by 16S AS (Continued) Mobiluncus curtisii 0 1 Gram Positive Anaerobe Morganella morganii 0 1 Gram Negative Facultative anaerobe Moryella indoligenes 0 1 Gram Positive Anaerobe Negativicoccus succinicivorans 1 1 Gram Negative Anaerobe Parvimonas micra 1 0 Gram Positive Anaerobe Peptoniphilus gorbachii 4 4 Gram Positive Anaerobe Peptoniphilus ivorii 2 3 Gram Positive Anaerobe Peptoniphilus lacrimalis 2 1 Gram Positive Anaerobe Peptoniphilus olsenii 2 0 Gram Positive Anaerobe Peptostreptococcus anaerobius 2 0 Gram Positive Anaerobe Peptostreptococcus stomatis 1 0 Gram Positive Anaerobe Porphyromonas asaccharolytica 3 2 Gram Negative Anaerobe Porphyromonas bennonis 2 1 Gram Negative Anaerobe Porphyromonas somerae 2 2 Gram Negative Anaerobe Porphyromonas uenonis 1 1 Gram Negative Anaerobe Prevotella bergensis 1 1 Gram Negative Anaerobe Prevotella buccalis 2 0 Gram Negative Anaerobe Prevotella corporis 1 0 Gram Negative Anaerobe Prevotella intermedia 1 0 Gram Negative Anaerobe Prevotella timonensis 1 1 Gram Negative Anaerobe Proteus myxofaciens 0 1 Gram Negative Anaerobe Pseudomonas indica 0 1 Gram Negative Aerobe Pseudomonas otitidis 0 1 Gram Negative Aerobe Psychrobacter lutiphocae 1 0 Gram Negative Aerobe Serratia grimesii 0 1 Gram Negative Facultative anaerobe Staphylococcus carnosus 1 0 Gram Positive Facultative anaerobe Staphylococcus chromogenes 2 0 Gram Positive Facultative anaerobe Staphylococcus devriesei 1 0 Gram Positive Facultative anaerobe Staphylococcus hominis 5 2 Gram Positive Facultative anaerobe Staphylococcus pettenkoferi 1 0 Gram Positive Facultative anaerobe Staphylococcus saprophyticus 0 1 Gram Positive Facultative anaerobe Stenotrophomonas pavanii 0 1 Gram Negative Aerobe Streptococcus agalactiae 2 0 Gram Positive Facultative anaerobe Streptococcus anginosus 0 1 Gram Positive Facultative anaerobe Streptococcus canis 1 0 Gram Positive Facultative anaerobe Streptococcus dysgalactiae 1 0 Gram Positive Facultative anaerobe Streptococcus infantarius 1 2 Gram Positive Facultative anaerobe Varibaculum cambriense 0 1 Gram Positive Anaerobe Veillonella dispar 1 0 Gram Negative Anaerobe Veillonella rogosae 1 0 Gram Negative Anaerobe the most frequently detected genera were Corynebac- indices indicated a increase in diversity on the recur- terium (5 samples), Peptoniphilus (4 samples) and rent ulcers (Fig. 3), while the new ulcers had higher Anaerococcus (4 samples). The relative abundance (%) levels of dominance. However, these differences were of each genus is displayed in Fig. 1 for new ulcers and not statistically significant (unpaired t-test, Shannon Fig. 2 for recurrent ulcers. Dominance and Diversity value p = 0.3287, Dominance value p = 0.1649). Smith et al. BMC Microbiology (2016) 16:54 Page 5 of 12

Peptostreptococcus Sample 18 Streptococcus Parvimonas Anaerococcus Sample 17 Peptoniphilus Corynebacterium Porphyromonas Sample 13 Prevotella Campylobacter Sample 12 Bulleidia Clostridium Bacteriodes Sample 9 Fusobacterium Haemophilus Negativococcus Wound Samples Sample 8 Finegoldia Actinobaculum Dermabacter Sample 7 Eggerthella Staphylococcus Sample 5 Kocuria Dialister Helcococcus Sample 4 Bilophila Veillonella 0 20406080100Psychrobacter Abundance (%) Others (<0.5% abundance) Fig. 1 Abundance (%) of bacterial genus within new diabetic foot ulcers

Alcaligenes Anaerococcus Sample 20 Pseudomonas Corynebacterium Peptoniphilus Finegoldia Sample 19 Actinomyces Dermabacter Leclercia Enterobacter Sample 15 Serratia Escherichia Stenotrophomonas Staphylococcus Sample 14 Moryella Porphyromonas Helcococcus Prevotella

Wound Samples Wound Sample 6 Campylobacter Negativicoccus Actinobaculum Streptococcus Sample 3 Enterococcus Gemella Proteus Haemophilus Sample 2 Granulicatella Morganella Mobiluncus 020406080100Varibaculum Others (<0.5% abundance) Abundance (%) Fig. 2 Abundance (%) of bacterial genus within recurrent diabetic foot ulcers Smith et al. BMC Microbiology (2016) 16:54 Page 6 of 12

Fig. 3 Dominance and diversity (Shannon Value) indices for both sets of ulcers

The majority of OTU’s residing in both new and re- Ulcer and clinical characteristics correlations current ulcers (over 67 %) were identified as Gram- Both HbA1c and the duration of the patient’s diabetes positiveorganisms.TheseweremostlyGram-positive had correlated with dominance (p = 0.0174) and diver- cocci, such as Staphylococcus, Streptococcus, Anaero- sity (p = 0.0168) values. A lower HbA1c value and coccus, Peptoniphilus and Finegoldia. However, the shorter duration of diabetes correlated with the Gram-positive rods, Corynebacterium, Clostridium and higher diversity (lower dominance statistic and higher Actinomyces were also frequently detected. In both Shannon value) within the ulcer. No other ulcer char- types of ulcers the most frequently identified Gram- acteristics, including predominant genera identified by negative organisms were Porphyromonas spp. In all 16S AS, number of OTUs, oxygen tolerance and bac- sixteen ulcers sampled the majority of species identi- terial morphology, had any correlation with the pa- fied by 16S AS were classed as facultative or strict tient’s characteristics. anaerobes. In newly formed ulcers, only one of the 94 OTU’s identified (1.06 %) was an aerobe and in recur- Discussion ring ulcers, 4 of the 73 OTU’s identified (5.48 %) Diabetic foot ulcers are a common complication of were aerobes. The results at OTU’s level are outlined poorly controlled diabetes and are a significant cause of in Table 2. Comparison of the bacterial classes and morbidity and hospitalisation in sufferers of this disease genera within each group showed a significant differ- [3]. These debilitating wounds heal slowly and in severe ence (p < 0.05) between the amount of Gammaproteo- cases, lower extremity amputation may be the only clin- bacteria in the two groups. There were no other ical option [1]. For many years the role of bacteria in significant differences. In principal component (PC) chronic wound healing was often overlooked, as ap- analysis (Fig. 4), PC1 was significantly different (p = proximately half of diabetic foot ulcers exhibit no clinical 0.0229) between the two groups once the outlier sam- evidence of infection [13]. However, many individuals ple (sample 13) was removed. However, we were un- with diabetes have an impaired inflammatory response able to determine which bacteria are contributing to and may not show the classical signs of infection in a this difference due to a lack of clear clusters in wound with a high microbial burden [14]. The concept graphical format. that non-healing in chronic wounds is associated with

Fig. 4 Principal Component Plots. New ulcers (black circles) and recurrent ulcers (red circles) were plotted with a and (b) without sample 13 Smith et al. BMC Microbiology (2016) 16:54 Page 7 of 12

bacterial load was introduced [5, 15], and bacterial col- microbiome of chronic wounds, only detecting approxi- onisation and proliferation within diabetic foot ulcers is mately 1 % of the inhabiting bacteria, which is biased by now believed to significantly retard wound healing. selective culture [16, 23]. Recent studies using molecular Therefore, a greater understanding of the microbiome of methods have confirmed that chronic wounds, including these chronic wounds is urgently required to help guide diabetic foot ulcers, have a polymicrobial nature instead the successful treatment [16]. Here we report for the of being colonised by a single species [24]. In this study, first time the complex microbiome of new and recurrent the number of OTU’s in a new ulcer samples ranged ulcers, demonstrating how diverse polymicrobial biofilm from 2 to 21, and in recurrent ulcers species ranged populations are common in diabetic patients. This is of from 6 to 17. There is growing evidence that, as with significance as there is growing evidence that polymicro- other persistent infections, the bacteria that reside bial interactions may synergise the pathogenic potential within chronic wounds grow within biofilm communities of one or other microorganism [17]. This has implica- [18, 19, 25]. This was supported by studies utilising mi- tions for patient management, as eradication of microor- croscopy that have shown that specimens from 60 % of ganisms could be important in controlling these chronic chronic wounds contained polymicrobial biofilm struc- wounds [7, 18, 19]. tures [6]. The presence of bacterial cells encased within In the clinic, infection is generally suspected based a biofilm may contribute to the chronicity of infection, on the presence of at least two classic signs of inflam- as biofilm-associated cells are notoriously recalcitrant. mation (erythema, warmth, tenderness, pain or indur- Poor penetration of the biofilm structure and extracellu- ation) or purulent secretions [20]. The standard lar matrix, nutrient limitation leading to slow growth method of diagnosis is by traditional laboratory cul- and phenotypic variants, protect the cells from the ef- ture of a sample taken directly from the wound. Our fects of antimicrobials and the host immune response study relied on swab sampling rather than wound bi- [7, 18]. The eradication of polymicrobial biofilms opsy because there was a desire to limit the use of in- within diabetic foot ulcers could key to resolving these vasive procedures. The presence of bacteria was chronic wounds. detected in ten of the twenty swab samples cultured. The analysis of the microbiome of this patient group All other samples produced no significant growth. showed that the most frequently identified genera were Standard laboratory reporting only provided basic in- the Gram-positive facultative anaerobes, Staphylococcus formation and did not readily identify the precise spe- and Corynebacterium. These organisms are part of the cies present in the majority of samples, though S. normal microbiota of healthy skin particularly in moist aureus was the most commonly isolated pathogen areas, such as the foot [26, 27]. Dowd et al. hypothesised (40 % of samples). It has been well documented in the concept that individual bacterial species may not be studies of diabetic foot ulcers that S. aureus is the able to maintain a pathogenic biofilm alone, but in a most commonly detected pathogen by laboratory cul- symbiotic polymicrobial community in a DFU, patho- ture [21, 22]. The frequent identification may be due genic biofilm may form [24]. Therefore, although these to the ability of staphylococci to grow under normal bacteria are normally commensals, they may be contrib- laboratory conditions when these methods often fail to uting to a pathogenic community. In our study, S. aur- identify slow-growing, fastidious or anaerobic organ- eus was not detected by 16S AS, instead all species isms [23]. This does not mean that these organisms identified were coagulase-negative staphylococci (CoNS). are an insignificant coloniser of chronic wounds. In a We confirmed the presence of S. aureus DNA in these retrospective study, it was found that 79 % of wounds samples using PCR, so while 16S AS is undoubtedly ro- sampledwereinfectedwithS. aureus [21]. More bust at the genus level results expressed at species level alarmingly, 30 % of these isolates were methicillin re- should be interpreted with some caution. This under- sistant S. aureus (MRSA). In this study 7 of the 8 lines the current problem with using data from 16S AS Staphylococcus isolates detected in our study were re- at the species level [28]. In the past, CoNS and Coryne- sistant to penicillin in culture, and 3 of these isolates bacterium spp. have been dismissed as contaminants of were positive for mecA by PCR. Therefore, culture- normal skin flora in diabetic foot ulcers, but several based methods still play an important role in patient studies have highlighted their importance as potential management, but do not necessarily give a true repre- pathogens. In the case of CoNS, there is a link between sentation of the pathogenic burden. the presence of these species in diabetic foot ulcers and Conventional culture techniques have a tendency to the incidence of osteomyelitis [29, 30]. Armstrong et al. produced false negative results, with over 37 % of sam- (1995) studied a predominantly diabetic patient group ples showing no signs of infection by culture alone. It is with osteomyelitis and found that 40 % of bone cultures now widely accepted that past reliance on standard cul- were positive for CoNS, 63 % of which were resistant to ture techniques has led to an underestimation of the the antibiotic methicillin [29]. A high incidence of Smith et al. BMC Microbiology (2016) 16:54 Page 8 of 12

Corynebacterium spp. has also been reported in studies bacteria includes Enterobacteriaceae and Pseudomona- using both culture and molecular tools to analyse the dales. Our findings correlate with a culture-based study bio-burden of diabetic foot ulcers [24, 31, 32], and com- that reported a high incidence of members of the bined with our findings highlights the importance of Pseudomonas, Enterococcus and Enterobacteriaceae CoNS and Corynebacterium spp. in relation to chronic groups in moderate to severe diabetic foot ulcers [22]. It wounds, particularly in individuals with diabetes who is well documented that Pseudomonas aeruginosa is fre- may have an impaired immune response [22, 29–32]. quently isolated from infected diabetic foot ulcers where In the biological niche of the diabetic ulcer the com- it is thought to play a role in severe tissue damage [37]. bination of necrotic tissue and low oxygen tension pro- Clinical isolates of this organism collected from chronic motes the proliferation of facultative or obligate wounds may also be multi-drug resistant which makes anaerobes [14, 33]. The majority of OTU’s identified in them very difficult to eradicate with antibiotic therapy our study conform to this logic, with obligate anaerobes [37]. In a longitudinal study of wound microbiota in an making up 65.9 % of the OTU’s in new ulcers and animal diabetic ulcer model it was reported that as time 56.1 % in recurrent ulcers. Only 1 OTU detected in new progressed there was a significant shift in bacterial type ulcers and 4 OTU’s in recurrent ulcers were aerobic or- from Furmicutes to genera including Enterobacter, ganisms. By culture our study was only able to identify which produced a corresponding decline in wound heal- anaerobes in 25 % samples tested but by 16S AS anaer- ing [26, 38]. This shift towards the presence of enteric- obes were found in 87.5 % of samples screened. The types of bacteria in the recurrent wound may be a result most frequently identified anaerobes in new ulcers were of self-colonisation from another body site. No other sig- in the genera Peptoniphilus (6/9 new ulcers), Anaerococ- nificant differences were found between the two groups, cus (5/9 new ulcers), Finegoldia (5/9 new ulcers), Por- including in diversity (Fig. 3). There was no difference phyromonas (4/9 new ulcers), and Prevotella (3/9 new between the two groups in bacteria with different oxygen ulcers). In recurring ulcers the most frequently identified tolerance or number of OTU. There was a significant anaerobes were also Finegoldia (5/7 recurring ulcers), difference in the number of Gram-positives in each Peptoniphilus (4/7 recurring ulcers), Anaerococcus (4/7 group, the recurrent group having less Gram-positive recurring ulcers), Porphyromonas (2/7 recurring ulcers), bacteria. When further broken down into morphological with the addition of Actinomyces (4/7 recurring ulcers), types this difference was no longer significant. This dif- which was not detected in new ulcers. The obligate an- ference in Gram positive/negative balance could be aerobes occurred as part of a polymicrobial microbiome accounted for by the difference in levels of Gammapro- in all cases, and in previous studies this has earned them teobacteria. Additionally PCA analysis (Fig. 4) indicated the name of ‘co-pathogens’, playing down their import- no formation of distinct groups. Further longitudinal in- ance [20]. The fact that anaerobes were detected in over vestigations monitoring the ulcer microbiome over time 87 % of the ulcers screened in our study suggests that will help determine if the presence of certain species are they may have a much more important role. Our data associated with the chronicity of these wounds and the agrees with three previous molecular studies of chronic inhibition of healing. wounds which discovered a high incidence of anaerobic As lower HbA1c levels and a shorter duration of dia- organisms including Anaerococcus, Finegoldia and Pep- betes correlated with higher diversity within the ulcer toniphilus [11, 24, 32, 34]. Although wounds are gener- this suggests that poorly glycaemic control and per- ally exposed to air, anaerobes may be able to survive if sistent diabetes causes a dominant bacterial species to they co-aggregate with facultative anaerobes or aerobe rise within the ulcer. In contrast, Gardner et al. found within a polymicrobial biofilm structure, which would that there was no link between high HbA1c levels and protect them from the harmful effects of oxygen and diversity, but instead poor glycaemic control was allow them to thrive [35]. The virulence of anaerobic or- linked to a higher abundance of Staphylococci and ganisms has also recently been highlighted in a study of Streptococci [11]. Finegoldia magna, found in 62.5 % of ulcers in our Optimal treatment of infection relies on accurate diag- study, which revealed that this pathogen produces an nosis of the microbes present and delivery of appropriate extracellular serine protease (SufA) to degrade collagen antimicrobial treatment. Failure to effectively treat the in the skin basement membrane [36]. Using such mech- infection in diabetic foot ulcers leads to progressive tis- anisms, anaerobes may be responsible for much of the sue damage, disrupted wound healing, and serious com- pathogenesis associated with chronic diabetic foot plications such as osteomyelitis [39]. Due to the reliance ulcers. on traditional laboratory culture, many clinics underesti- The levels of Gammaproteobacteria in the two groups mate wound flora, as highlighted by our study and were also found to be significantly different, with the re- others, who compared laboratory culture with molecular current group having a higher level. This class of methods, discovering that in 45 % of cases an Smith et al. BMC Microbiology (2016) 16:54 Page 9 of 12

inappropriate antibiotic was prescribed [33]. Clinical development of a unique microbiome within each Practice Guidelines (2012) recommend that only dia- diabetic foot ulcer. betic foot ulcers with clinical signs of infection are 3) 16S AS cannot currently identify all organisms treated with antibiotics due to the adverse effects, fi- reliably at a species level and this must be taken into nancial costs and increasing risk of antibiotic resistance consideration when using this technique to [20]. There have been reports that individuals with dia- characterise the microbiome of an infection site. betic foot ulcers treated with antibiotics, in the absence of clinical signs of infection, exhibited significantly im- Greater understanding of the diabetic foot ulcer proved rates of healing in comparison to those who did microbiome will help guide new strategies to effectively not undergo antibiotic therapy [40]. The clinical guide- control the growth of polymicrobial biofilms and im- lines do recommend that patients with signs of mild to prove healing, directly benefiting patients suffering from moderate infection receive antibiotic therapy targeting these debilitating wounds. aerobic Gram-positive cocci [20]. In our study, the majority of bacteria detected in both new and recurring Methods ulcers were Gram-positive facultative or obligate anaer- Subject selection obes, which would require antibiotics with a wider The protocol for this study was approved by the spectrum of activity to successfully resolve the infec- National Research Ethics Committee (London, UK) (Study tion. Metronidazole is the drug of choice to effectively Reference 13/LO/1509) and the Research and Development treat a range of infections caused by anaerobic bacteria Office (NHS Ayrshire & Arran). Twenty patients with dia- and it may have an important role to play in the man- betes attending the Diabetes/Podiatry clinic at University agement of chronic diabetic foot ulcer infection [41], Hospital Ayr (NHS Ayrshire & Arran) were selected for this though the evidence of its effectiveness in these infec- study and written consent was obtained. Patients were tions is questionable [42, 43]. The clinical guidelines excluded if they had received antibiotic treatment in also stress that definitive therapy be based on obtaining the past 3 months. All study participants were Caucasian, appropriate culture results from the clinical laboratory, diagnosed with Type 2 diabetes and the group comprised however we have shown that the use of clinical culture of 18 males and 2 females with an age range of 51 to alone may severely underestimate the microbial load 86 years. Ten subjects were selected with a new ulcer within ulcers and cause a delay in appropriate treat- (within 3 days of ulcer formation) and ten subjects were ment. Culture independent molecular methods, such as selected with a recurring ulcer (any secondary ulcer re- PCR, have revolutionized many areas of clinical micro- gardless of location) [44]. biology [34] and it is crucial that these are applied to the routine monitoring of infection in diabetic foot ul- Swab sample collection cers to diagnose the pathogens responsible and guide During a routine visit to the podiatry clinic patients had appropriate treatment, focussing on disruption of the their wound dressing changed and ulcer examined. At biofilm. this time, standard protocols were followed to sample the total surface of the foot ulcer with a Regular Conclusions FLOQSwab™ (COPAN, Brescia, Italy), which was directly Due to the minimal differences found in this study be- placed into a tube containing 2 ml of sterile PBS and im- tween the bacterial colonisation of new and recurrent, mediately placed on ice, for DNA isolation. A second there is no need to treat with more severe methods for a swab (Sterilin, Thermo Scientific, UK) was then used to secondary ulcer, even though it is associated with a sample the same area of the wound and sent to the clin- higher risk of infection [20]. Due to minimal differences ical laboratory at University Hospital, Crosshouse (NHS between the bacterial colonisation of new and recurrent Ayrshire & Arran) for standard aerobic and anaerobic ulcers, there is no need to treat with more severe culture. Swab sampling was used as the preferred methods for a secondary ulcer, even though it is associ- method of sampling in this study as it is a less invasive ated with a higher risk of infection [20]. alternative to ulcer biopsy and limits the risk of introdu- Our study produced three key findings: cing infection in this vulnerable group of patients. Two control swabs were also was taken of the healthy skin on 1) The complexity of the bacterial population present the unaffected foot of each patient. Hospital labs gave in diabetic foot ulcers is much greater than would results in the form of presence or absence of species or be expected from culture studies alone. groups of bacteria. One swab was used for DNA isola- 2) There is no significant difference in the bacterial tion and the other was assessed by aerobic and anaer- populations in new and recurrent ulcers, suggesting obic culture to discount any bacterial strains, which may that each wound provides a blank canvas for the be part of the patient’s normal healthy skin flora. Smith et al. BMC Microbiology (2016) 16:54 Page 10 of 12

Information on current HbA1c levels and the duration hit [47]. Chimeric clusters were removed using of the diabetes were also taken from the patient. A UCHIME [48], and unique clusters were subjected to Standard PCR was performed to determine the presence BLASTn [49] analysis. Good quality and unique 16S or absence of methicillin-sensitive Staphylococcus aureus rDNA sequences were used as a reference database to (MSSA) and methicillin-resistant S. aureus (MRSA) assign operational taxonomic unit (OTU) status to the using primers and conditions previously described by clusters. Classification of OTU clusters and the number our group [45]. Primer sequences were as follows: of reads within were consolidated to compute the relative abundance of each species within each sample. S. aureus: F- ATTTGGTCCCAGTGGTGTGGGTAT, R-GCTGTGACAATTGCCGTTTGTCGT, Statistical analyses S. aureus mecA primers: F- AACCACCCAATTTGTC Spearman’s rank correlation was used to assess correl- TGCC, ation between culture and PCR results using GraphPad R- TGATGGTATGCAACAAGTCGTAAA. Prism software version 6. OTU datasets were reduced by log2 transformation so DNA isolation and purification as to carry out principal component analysis (PCA) and The DNA was isolated from bacterial cells carried on diversity statistics (Shannon diversity index and Domin- each FLOQSwab™ within 2 h of clinical sample collec- ance index); the analysis was carried out using PAST tion. Bacterial cells were released from FLOQSwabs™ software [50]. into the 2 ml of PBS by sonicating the samples in a An unpaired t-test was applied to compare diversity water bath (Fisherbrand, ThermoFisher Scientific Inc., statistics and ulcer bacterial characteristics (oxygen tol- Loughborough, UK) at 35 kHz for 3 × 20s, and vortexing erance, bacterial morphology, and number of OTU) for 10s. The samples were centrifuged for 5 min at 5223 using GraphPad Prism® software version 6. PCA was × g to pellet the cells. DNA was then isolated from each used to reduce the dimensionality of the OTU dataset. A sample using bead-beating combined with a QIAamp scree plot was used to determine how many components DNA Mini Kit (Qiagen, UK), according to the manufac- emerged. No distinct clusters appeared between the two turer’s instructions. DNA was stored at −20 °C until re- groups. Sample 13 was distant from the other samples quired. The concentration and integrity of purified DNA and may have been skewing the data. Therefore analysis was measured by NanoDrop analysis (Thermo Scientific, was repeated without Sample 13. To determine if dis- UK). It was our initial intention in this study to analyse tinct clusters formed for each group on the PCA plots, the microbiome of the healthy control skin of each indi- new variables were created for each principle component vidual to compare with the microbiome of each ulcer. by using the factor loadings as regression coefficients, However, the concentration of DNA isolated from con- producing a score for each sample. These scores were trol samples was extremely low (in the range 0.5- then used as outcome variables to compare between 1.27 ng/μl) and was much lower than the concentration groups. required for sequencing (20 ng/μL), therefore these sam- The contribution of each bacterial class and genera ples were removed from the study and control samples that represented over 1 % of the group was calculated in were analysed by traditional laboratory culture only. terms of proportion to the overall sample, percentages were log transformed and an unpaired t-test was used to 16S amplicon sequencing (16S AS) and sequence analysis compare new and recurrent groups. 16S AS was performed by GATC Biotech AG (Konstanz, Characteristics of the ulcers (bacterial morphology, Germany). DNA samples (total volume 30 μl [20 ng/μL]) Gram type, oxygen tolerance, diversity levels, and com- were submitted for initial PCR amplification using mon genera) were correlated to clinical aspects of the primers for the V4 region of the 16S rDNA gene patient (HbA1c levels and duration of diabetes) using (515 F- AGAGTTTGATCCTGGCTCAG and 806R two-tailed Spearman’s correlation in GraphPad Prism. ATTACCGCGGCTGCTGG), producing a 253 bp product for sequencing. The Illumina HiSeq 2500 plat- Availability of data and materials form (Illumina Inc., San Diego, USA) was used to pro- Raw data sets are available from the following website: duce paired-end sequence reads using cyclic reversible http://www.ncbi.nlm.nih.gov/ bioproject/304940. chain termination chemistry. The base call of each sequence read was inspected and filtered for quality. Se- Abbreviations quences below the quality threshold or less than 137 bp 16S AS: 16S amplicon sequencing; CoNS: coagulase negative staphylococci; OTU: operational taxonomic unit. in length were removed. Sequence read pairs were merged using FLASh [46] and compressed based on Competing interests 99 % similarity, using the clustering program CD-HIT- None of the authors have any conflict of interest in publishing this work. Smith et al. BMC Microbiology (2016) 16:54 Page 11 of 12

Authors’ contributions 18. Stewart PS. Mechanisms of antibiotic resistance in bacterial biofilms. Int J KS participated in the study design, was responsible for sample collection, Med Microbiol. 2002;292(2):107–13. carried out the experimental studies, and was responsible for writing the 19. Percival SL, Emanuel C, Cutting KF, Williams DW. Microbiology of the skin manuscript. AC and AMB assisted with the collection of samples and their and the role of biofilms in infection. Int Wound J. 2012;9(1):14–32. clinical interpretation. JB organized and submitted the sequencing data. EMT 20. Lipsky BA, Berendt AR, Cornia PB, Pile JC, Peters EJ, Armstrong DG, Deery carried out the experimental studies, and along with LEO carried out HG, Embil JM, Joseph WS, Karchmer AW. 2012 infectious diseases society of statistical analyses. WGM, GR and CW participated in study design, analysis America clinical practice guideline for the diagnosis and treatment of and were responsible for writing and submission of the final manuscript. All diabetic foot infections. Clin Infect Dis. 2012;54(12):e132–73. authors contributed, read and approved the manuscript. 21. Dang C, Prasad Y, Boulton A, Jude E. Methicillin‐resistant Staphylococcus aureus in the diabetic foot clinic: a worsening problem. Diabet Med. – Acknowledgements 2003;20(2):159 61. We thank Claire MacKenzie for her assistance in the collection of clinical 22. Citron DM, Goldstein EJ, Merriam CV, Lipsky BA, Abramson MA. Bacteriology samples. of moderate-to-severe diabetic foot infections and in vitro activity of antimicrobial agents. J Clin Microbiol. 2007;45(9):2819–28. Author details 23. Kaeberlein T, Lewis K, Epstein SS. Isolating” uncultivable” microorganisms 1School of Health, Nursing and Midwifery, Institute of Healthcare Associated in pure culture in a simulated natural environment. Science. Infection, University of the West of Scotland, Barbour Building, Paisley 2002;296(5570):1127–9. Campus, Paisley PA1 2BE, UK. 2NHS Ayrshire and Arran, University Hospital 24. Dowd SE, Wolcott RD, Sun Y, McKeehan T, Smith E, Rhoads D. Polymicrobial Ayr, Ayr, UK. 3Infection and Immunity Research Group, Glasgow Dental nature of chronic diabetic foot ulcer biofilm infections determined using School, School of Medicine, College of Medical, Veterinary and Life Sciences, bacterial tag encoded FLX amplicon pyrosequencing (bTEFAP). PLoS One. University of Glasgow, Glasgow, UK. 2008;3(10), e3326. 25. Peters BM, Jabra-Rizk MA, Graeme A, Costerton JW, Shirtliff ME. Received: 20 August 2015 Accepted: 7 March 2016 Polymicrobial interactions: impact on pathogenesis and human disease. Clin Microbiol Rev. 2012;25(1):193–213. 26. Grice EA, Segre JA. The skin microbiome. Nat Rev Microbiol. 2011;9(4):244–53. References 27. Grice EA, Kong HH, Conlan S, Deming CB, Davis J, Young AC, Bouffard GG, 1. Singh N, Armstrong DG, Lipsky BA. Preventing foot ulcers in patients with Blakesley RW, Murray PR, Green ED. Topographical and temporal diversity of – diabetes. Jama. 2005;293(2):217–28. the human skin microbiome. Science. 2009;324(5931):1190 2. 2. Bakker K, Apelqvist J, Schaper N. Practical guidelines on the management 28. Ghebremedhin B, Layer F, König W, König B. Genetic classification and and prevention of the diabetic foot 2011. Diabetes Metab Res Rev. distinguishing of staphylococcus species based on different partial gap, 16S 2012;28(S1):225–31. rRNA, hsp60, rpoB, sodA, and tuf gene sequences. J Clin Microbiol. – 3. Alexiadou K, Doupis J. Management of diabetic foot ulcers. Diabetes 2008;46(3):1019 25. Therapy. 2012;3(1):1–15. 29. Armstrong DG, Lanthier J, Lelievre P, Edelson GW. Methicillin-resistant 4. Boulton AJ, Vileikyte L, Ragnarson-Tennvall G, Apelqvist J. The global burden coagulase-negative staphylococcal osteomyelitis and its relationship to of diabetic foot disease. Lancet. 2005;366(9498):1719–24. broad-spectrum oral antibiosis in a predominantly diabetic population. – 5. Robson MC. Wound infection: a failure of wound healing caused by an J Foot Ankle Surg. 1995;34(6):563 6. imbalance of bacteria. Surg Clin North Am. 1997;77(3):637–50. 30. Loïez C, Wallet F, Pischedda P, Renaux E, Senneville E, Mehdi N, Courcol RJ. “ ” 6. James GA, Swogger E, Wolcott R, Secor P, Sestrich J, Costerton JW, Stewart First case of osteomyelitis caused by Staphylococcus pettenkoferi . J Clin – PS. Biofilms in chronic wounds. Wound Repair Regen. 2008;16(1):37–44. Microbiol. 2007;45(3):1069 71. 7. Mah T-FC, O'Toole GA. Mechanisms of biofilm resistance to antimicrobial 31. Bessman AN, Geiger PJ, Canawati H. Prevalence of corynebacteria in – agents. Trends Microbiol. 2001;9(1):34–9. diabetic foot infections. Diabetes Care. 1992;15(11):1531 3. 8. Hex N, Bartlett C, Wright D, Taylor M, Varley D. Estimating the current and 32. Gontcharova V, Youn E, Sun Y, Wolcott RD, Dowd SE. A comparison of future costs of type 1 and type 2 diabetes in the UK, including direct bacterial composition in diabetic ulcers and contralateral intact skin. The health costs and indirect societal and productivity costs. Diabet Med. open microbiology journal. 2010;4:8. 2012;29(7):855–62. 33. Dunyach-Remy C, Cadière A, Richard J-L, Schuldiner S, Bayle S, Roig B, Sotto 9. Doupis J, Veves A. Classification, diagnosis, and treatment of diabetic A, Lavigne J-P. Polymerase chain reaction—denaturing gradient gel foot ulcers. Wounds: a compendium of clinical research and practice. electrophoresis (PCR–DGGE): a promising tool to diagnose bacterial 2008;20(5):117–26. infections in diabetic foot ulcers. Diabetes Metab. 2014;40(6):476–80. 10. Oates A, Bowling FL, Boulton AJ, McBain AJ. Molecular and culture-based 34. Sibley CD, Peirano G, Church DL. Molecular methods for pathogen and assessment of the microbial diversity of diabetic chronic foot wounds and microbial community detection and characterization: current and potential contralateral skin sites. J Clin Microbiol. 2012;50(7):2263–71. application in diagnostic microbiology. Infect Genet Evol. 2012;12(3):505–21. 11. Gardner SE, Hillis SL, Heilmann K, Segre JA, Grice EA. The neuropathic 35. Sun Y, Smith E, Wolcott R, Dowd S. Propagation of anaerobic bacteria diabetic foot ulcer microbiome is associated with clinical factors. Diabetes. within an aerobic multi-species chronic wound biofilm model. J Wound 2013;62(3):923–30. Care. 2009;18(10):426–31. 12. Consortium HMP. A framework for human microbiome research. Nature. 36. Murphy EC, Mörgelin M, Reinhardt DP, Olin AI, Björck L, Frick IM. 2012;486(7402):215–21. Identification of molecular mechanisms used by finegoldia magna to 13. Lavery LA, Armstrong DG, Wunderlich RP, Mohler MJ, Wendel CS, Lipsky BA. penetrate and colonize human skin. Mol microbiol. 2014;94(2):403–17. Risk factors for foot infections in individuals with diabetes. Diabetes Care. 37. Sivanmaliappan TS, Sevanan M. MAntimicrobial susceptibility patterns of 2006;29(6):1288–93. Pseudomonas aeruginosa from diabetes patients with foot ulcers. Int j 14. Bjarnsholt T, Kirketerp‐Møller K, Jensen PØ, Madsen KG, Phipps R, Krogfelt K, microbiol. 2011;2011:4. Høiby N, Givskov M. Why chronic wounds will not heal: a novel hypothesis. 38. Grice EA, Snitkin ES, Yockey LJ, Bermudez DM, Liechty KW, Segre JA, Mullikin Wound Repair Regen. 2008;16(1):2–10. J, Blakesley R, Young A, Chu G. Longitudinal shift in diabetic wound 15. Xu L, McLennan SV, Lo L, Natfaji A, Bolton T, Liu Y, Twigg SM, Yue DK. microbiota correlates with prolonged skin defense response. Proc Natl Acad Bacterial load predicts healing rate in neuropathic diabetic foot ulcers. Sci. 2010;107(33):14799–804. Diabetes Care. 2007;30(2):378–80. 39. Lipsky BA. Treating diabetic foot osteomyelitis primarily with surgery or 16. Martin JM, Zenilman JM, Lazarus GS. Molecular microbiology: new antibiotics: have we answered the question? Diabetes Care. 2014;37(3):593–5. dimensions for cutaneous biology and wound healing. J Investig Dermatol. 40. Edmonds M, Foster A. The use of antibiotics in the diabetic foot. Am J Surg. 2010;130(1):38–48. 2004;187(5):S25–8. 17. Stacy A, Everett J, Jorth P, Trivedi U, Rumbaugh KP, Whiteley M. Bacterial 41. Löfmark S, Edlund C, Nord CE. Metronidazole is still the drug of choice for fight-and-flight responses enhance virulence in a polymicrobial infection. treatment of anaerobic infections. Clin infect dis. 2010;50(Supplement 1): Proc Natl Acad Sci U S A. 2014;111(21):7819–24. S16–23. Smith et al. BMC Microbiology (2016) 16:54 Page 12 of 12

42. Aherrao N, Shahi SK, Dwivedi A, Kumar A, Gupta S, Singh SK. Detection of anaerobic infection in diabetic foot ulcer using PCR technique and the status of metronidazole therapy on treatment outcome. Wounds : a compendium of clinical research and practice. 2012;24(10):283–8. 43. Clay PG, Graham MR, Lindsey CC, Lamp KC, Freeman C, Glaros A. Clinical efficacy, tolerability, and cost savings associated with the use of open-label metronidazole plus ceftriaxone once daily compared with ticarcillin/ clavulanate every 6 h as empiric treatment for diabetic lower-extremity infections in older males. Am J Geriatr Pharmacother. 2004;2(3):181–9. 44. Peters EJ, Armstrong DG, Lavery LA. Risk factors for recurrent diabetic foot ulcers: site matters. Diabetes Care. 2007;30(8):2077–9. 45. O'Donnell LE, Smith K, Williams C, Nile CJ, Lappin DF, Bradshaw D, Lambert M, Robertson DP, Bagg J, Hannah V et al. Dentures are a reservoir for respiratory pathogens Journal of Prosthodontics 2015; doi:10.1111/jopr.12342. 46. Magoč T, Salzberg SL. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics. 2011;27(21):2957–63. 47. Li W, Godzik A. Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics. 2006;22(13):1658–9. 48. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. UCHIME improves sensitivity and speed of chimera detection. Bioinformatics. 2011;27(16):2194–200. 49. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol. 1990;215(3):403–10. 50. Hammer OHD, Ryan PD. PAST: paleontological statistics software package for education and data analysis. Palaeontol Electron. 2001;4(1):9.

Submit your next manuscript to BioMed Central and we will help you at every step:

• We accept pre-submission inquiries • Our selector tool helps you to ﬁnd the most relevant journal • We provide round the clock customer support • Convenient online submission • Thorough peer review • Inclusion in PubMed and all major indexing services • Maximum visibility for your research

Submit your manuscript at www.biomedcentral.com/submit